unexpected families loaded in database

[sevyharris]

Bug Description

I'm trying to load a blank/mostly empty RMG database, but when I specify kinetics_families=[], it doesn't load a database with zero families. It turns out, you have to specify kinetics_families='none'

How To Reproduce

import rmgpy.data.rmg
database = rmgpy.data.rmg.RMGDatabase()
database.load(
    path=rmgpy.settings['database.directory'],
    thermo_libraries=['primaryThermoLibrary'],
    transport_libraries=[],
    reaction_libraries=[],
    seed_mechanisms=[],
    kinetics_families=[],
    kinetics_depositories=[],
    depository=False,
)

# Prepare the database by loading training reactions but not averaging the rate rules
for familyLabel, family in database.kinetics.families.items():
    if not family.auto_generated:
        family.add_rules_from_training(thermo_database=database.thermo)
        family.fill_rules_by_averaging_up(verbose=True)

Expected Behavior

I would expect an empty list to give me the same behavior as 'none' i.e. a database with zero families. Instead, it seems to be loading all the families. Or at least a large number of families.

Installation Information

• OS: Ubuntu 24.04.4 LTS
• Installation method: source
• RMG version information:
  • RMG-Py: 708fecd
  • RMG-database: ecb79707121726295363f92846cfdf88c4d7311d

Additional Context

This is a really petty thing to complain about, but there's a downstream problem where loading the database like this includes surface families, which require a surface library to do the data tree averaging or else you'll get this error that I want to say Jonathan had also run into for some of the RMG website stuff:

---------------------------------------------------------------------------
DatabaseError                             Traceback (most recent call last)
Cell In[28], [line 4](vscode-notebook-cell:?execution_count=28&line=4)
      2 for familyLabel, family in database.kinetics.families.items():
      3     if not family.auto_generated:
----> [4](vscode-notebook-cell:?execution_count=28&line=4)         family.add_rules_from_training(thermo_database=database.thermo)
      5         family.fill_rules_by_averaging_up(verbose=True)

File ~/rmg/RMG-Py/rmgpy/data/kinetics/family.py:1266, in KineticsFamily.add_rules_from_training(self, thermo_database, train_indices)
   1264     product.molecule[0].clear_labeled_atoms()
   1265     product.generate_resonance_structures()
-> [1266](https://file+.vscode-resource.vscode-cdn.net/home/moon/uncertainty_estimator/uncertainty_tool_dev/demos/quadrature/~/rmg/RMG-Py/rmgpy/data/kinetics/family.py:1266)     product.thermo = thermo_database.get_thermo_data(product, training_set=True, metal_to_scale_to=metal)
   1267 # Now that we have the thermo, we can get the reverse k(T)
   1268 item.kinetics = data

File ~/rmg/RMG-Py/rmgpy/data/thermo.py:1295, in ThermoDatabase.get_thermo_data(self, species, metal_to_scale_to, training_set)
   1293 if species.contains_surface_site():
   1294     try:
-> [1295](https://file+.vscode-resource.vscode-cdn.net/home/moon/uncertainty_estimator/uncertainty_tool_dev/demos/quadrature/~/rmg/RMG-Py/rmgpy/data/thermo.py:1295)         thermo0 = self.get_thermo_data_for_surface_species(species)
   1296         metal_to_scale_from = self.adsorption_groups.split('adsorption')[-1]
   1297         if metal_to_scale_from != metal_to_scale_to:

File ~/rmg/RMG-Py/rmgpy/data/thermo.py:1572, in ThermoDatabase.get_thermo_data_for_surface_species(self, species)
   1569         return species.thermo.get_enthalpy(298.0)
   1571 if species.is_surface_site():
...
   1574 logging.debug("Trying to generate thermo for surface species using first of %d resonance isomer(s):",
   1575               len(species.molecule))
   1577 resonance_data = []

DatabaseError: Can't estimate thermo of vacant site. Should be in library (and should be 0).